Dynamic Style Analysis of Hedge Funds and Kalman Smoother

Dr. Leonid Shvartser (speaker)
Israеl, TSG Advanced Systems Ltd
Michael Markov
USA, MPI International Inc
Dr. Olga Krasotkina
Russia, Moscow State University
Prof. Vadim Mottl’
Russia, Computing Center of the Russian Academy of Sciences
Presented comparative research was done by L. Shvartser and M. Markov; other two authors, O. Krasotkina and V. Mottl’, played critical role in DSA method development
Dynamic Style Analysis (DSA) is a method of restoration of strategy of a hedge fund on a given time interval: to estimate the fractions of the portfolio invested in different assets at every time stamp during the investigated period. The Flexible Least Squares (FLS) method solves this problem. In this work we show the correspondence between the two state estimators, FLS and Kalman Smoother, and analyze the merits and disadvantages of each. It is shown that the unknown smoothing parameter in DSA is a combination of the unknown measurement and model noises in the State Space Model. Analytical (non-combinatorial) cross-validation methods for estimation of the smoothing parameters and their comparison were developed. Different approaches to detection of the structural changes in the DSA models are analyzed: one of the approaches uses both forward and backward passes of Kalman Smoother but doesn’t use the estimation and the prediction errors; the second one uses only the forward pass which is the Kalman Filter but it finds the structural changes using pure statistical testing. The different approaches to the Dynamic Style Analysis of hedge funds are illustrated on real data benchmarks.